BCRAM: A Social-Network-Inspired Breast Cancer Risk Assessment Model

Abstract

The pathogenesis of breast cancer is not the same in all countries and regions; therefore, some existing breast cancer risk assessment models are not well adapted to all countries and regions, including China. This paper puts forward a new model named BCRAM (a social-network-inspired breast cancer risk assessment model) that depends on epidemiological factors, which is more adaptive to the populous country like China than those models based on gene. The model utilizes the similarities among epidemiological factors to construct a breast cancer high-risk group, the members of which have high similarity with breast cancer patients. Then, three tests based on real data are used to determine the assessment value of BCRAM. The AUC of BCRAM is 0.785, which is larger than that of the classic Gail model, a modified Gail model, the Tyrer–Cuzick model, and the Liu–Yu model for Chinese women. F-Measure value is 0.696, which is the largest among those of all models. Moreover, follow-up data are used to demonstrate that the model can give early warning to a high proportion of patients discovered to have breast cancer in the future. Therefore, the model is meaningful for the prevention and control of breast cancer. And the unique design of the method for selecting risk factors related to breast cancer results in our model having good generality, and it can be generalized to other countries and regions.